<html> <head> <base href="https://bugs.freedesktop.org/" /> </head> <body> <div> <a class="bz_bug_link bz_status_NEW " title="NEW - Flickering Shadows in The Talos Principle" href="https://bugs.freedesktop.org/show_bug.cgi?id=93594#c14">Comment # 14</a> on <a class="bz_bug_link bz_status_NEW " title="NEW - Flickering Shadows in The Talos Principle" href="https://bugs.freedesktop.org/show_bug.cgi?id=93594">bug 93594</a> from <a class="email" href="mailto:sroland@vmware.com" title="Roland Scheidegger <sroland@vmware.com>"> Roland Scheidegger</a> <pre>(In reply to Marek Olšák from <a href="show_bug.cgi?id=93594#c13">comment #13</a>) > (In reply to Roland Scheidegger from <a href="show_bug.cgi?id=93594#c12">comment #12</a>) > > Due to d3d10 being different there, I wouldn't be surprised if other apps > > make the same mistake. > > For performance, it's better to execute KILL as soon as possible, because it > allows us to skip texture loads and possibly even input loads. Certainly, which is probably why it is undefined behavior in glsl. My guess is the d3d10 drivers will check dependency chains and see if the result may be needed later for derivatives, and if it is only kill whole quads immediately or something like that. Whereas in GL it would be the responsibility of the shader to do the actual kill only when the pixels aren't needed any longer for calculating derivatives (though I suppose that means it potentially wouldn't early discard whole quads neither).</pre> </div> <hr> You are receiving this mail because: <ul> <li>You are the assignee for the bug.</li> </ul> </body> </html>